JP3356519B2 - Document information retrieval device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document information retrieval apparatus for retrieving a desired document from a document database storing a plurality of documents.

[0002]

2. Description of the Related Art In recent years, with the spread of word processors and computers, electronic news and e-mail have been put into practical use. Conventionally, information that could only be obtained on paper was published by CDROM (dictionary, encyclopedia, etc.) Computerization is progressing,
It is expected that the digitization of such documents will accelerate in the future.

[0003] With the digitization of such documents, an information retrieval system for efficiently retrieving a desired document from a large number of documents has been used in order to effectively use documents stored in advance.

[0004] As a conventional document information search system, there is known a system in which keywords are combined with logical operators such as AND, OR, and NOT to search for a document containing the keywords.

However, in such a conventional document information search system, since only the number of search results is displayed or only the title of the search result is displayed as the display of the search result, In order for the user to verify whether the result is the desired one, the user had to read the entire text of each retrieved document.

In a conventional document information retrieval system,
In accordance with the specification of a search command such as a keyword, the titles of the search result documents are simply displayed side by side in a certain order (for example, in the order of documents similar to the keyword). For this reason, the relative relationship between the retrieved documents and the degree of similarity with the retrieval command could not be felt by the user, and it was difficult for the user to intuitively perceive the validity of the retrieval result.

[0007] Furthermore, in the search methods known so far, only documents can be searched as one unit. For this reason, there is a problem in that the contents to be searched may be mixedly searched for without any distinction between the part described in the background and the part described as the conclusion. That is, documents described from all viewpoints can be obtained as a search result as a whole. For example, if you are n’t interested in “what has been done in the past,”
A document matched with the search command in the part described in the background is unnecessary. However, the conventional document information search system searches such a document without discriminating other documents, so that there is a great harm to the user in determining the validity of the search result. There was a problem of becoming.

[0008]

As described above, in the conventional document information search system, in order for the user to determine whether or not the search result is a desired document, the full text of each document is required. , And it takes a lot of time to obtain a desired document.

Therefore, a method of reducing the trouble of having to read the whole sentence by displaying only a part of the searched result has been considered. However, in this case, the displayed part is replaced by another part in the document. If it is not clear what the relationship is, it cannot be determined whether the document is a desired document. For example, if a part that describes background information is displayed for one document and a part that describes conclusions is displayed for another document, it is not possible to overlook from a unified viewpoint, so any document is necessary Is difficult to determine. Therefore, in order to understand the position of the displayed part with respect to the whole sentence, the whole sentence must be read, and as a result, labor is not reduced.

In addition, a summary sentence of each stored document is manually created, stored in association with the original document, and displayed in a search result to display the summary sentence. It is also considered to reduce the labor of verification by the elderly. However, even in this case, it takes a lot of time to manually create a summary sentence, and it cannot be applied unless the database has a very large use scale. Also,
So far, there are databases in which summaries of individual documents are not prepared, and manually summarizing them requires a great deal of cost and effort. Furthermore, since the summary sentence created by hand is created for the general public, it is not always summarized from a viewpoint suitable for the purpose of the user. For this reason, there is a problem that a user sometimes misses a necessary document when determining whether or not the document is a desired document in some cases because the sentence is out of focus.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and provides a document information system capable of quickly judging the validity of a document search result and smoothly performing a series of search operations for understanding individual documents. It is an object to provide a search device.

[0012]

SUMMARY OF THE INVENTION According to the present invention, there is provided a document information retrieval apparatus for retrieving a desired document from documents stored in a database, comprising: a document data storage unit for storing the document; An input unit for inputting a natural language sentence, an input analyzing unit for analyzing an input from the input unit, converting the input into a plurality of different viewpoints, and converting the input into a search command, and a search target stored in the document data storage unit. And a search unit that calculates a distance between the document and the search command for each viewpoint and obtains a document with a short distance.

[0013]

As a result, according to the present invention, when the user designates with a keyword or a natural language sentence, the input can be analyzed to search for a document from a plurality of different viewpoints.
Since the degree of similarity with the search command can be displayed based on these multiple viewpoints, the user can intuitively judge the validity of the search result from this display result and understand individual documents. A series of operations can be performed efficiently, and efficient document search can be realized even when there are many search documents.

That is, the result is displayed in a multidimensional space based on the distance between the search command and the document obtained for each of a plurality of viewpoints. To see if each document is similar.
Users will be able to understand based on the list.
And, since the document displayed at a close position in the multidimensional display has similar contents, it is possible to easily analogize the contents of the documents in the vicinity by reading only a part of the documents, By providing a function for easily browsing the contents of the document from the display of the search result, the trouble of reading the document can be drastically reduced. Further, even when there are many documents as a result of the search, the documents can be easily selected by clustering the documents.

[0015]

Embodiments of the present invention will be described below with reference to the drawings.

(First Embodiment) FIG. 1 shows a schematic configuration of a first embodiment. In the figure, reference numeral 101 denotes a central processing unit, a storage unit 102 such as a semiconductor memory, a magnetic disk, or an optical disk for storing programs and data in the central processing unit 101; Means 104 such as a liquid crystal display or a plasma display for performing
And input means 10 such as a keyboard or a mouse for inputting a user's search command via input controller 105.
6 are connected.

FIG. 2 shows a functional configuration of the first embodiment. An input unit 201, an input analysis unit 202, a search unit 203,
It comprises a history management unit 204, a search history storage unit 205, a document data storage unit 206, a search result display unit 207, a search history display unit 208, and a browse unit 209.

In this case, the input unit 201 inputs a keyword instructing a search or a sentence composed of a natural language sentence, and inputs it via the data lines 210 and 211 to the input analysis unit 20.
2, the user input is converted into a search command, the search node ID is input from the history management unit 204 via the data line 213, and the search node ID and the search command are paired to the data line 2.
12 to the history management unit 204 and to the search unit 203 via the data line 214.

The input analysis unit 202 receives a user's input from the input unit 201 via the data line 210, converts the input into a search command, and outputs the search command to the input unit 201 via the data line 211. .

The search unit 203 includes a search command and a search node I.
D is input from the input unit 201 via the data line 214, and the document data stored in the document data storage unit 206 is referenced via the data line 218 to search for a related document set. Is output to the history management unit 204 via the data line 215 in combination with the search node ID, and is output to the search result display unit 208 via the data line 216.

The search result display unit 207 inputs the search result via the data line 216, displays a document set of the search result as a multi-dimensional display, and displays the document ID of the document selected in the multi-dimensional display on the data line 227. Is output to the browsing unit 209 via.

The browsing unit 209 refers to the document data stored in the document data storage unit 206 via the data line 219 according to the document ID input from the search result display unit 207 via the data line 227, and Display the contents of the document.

The history management unit 204 includes a pair of a search command and a search node ID input via the data line 212, a search result input via the data line 215, and a search node I.
According to the pair with D, the search history is stored in the search history storage unit 205 via the data line 223, the search history is referred to via the data line 222, and the search history information is output to the search history display unit 207 via the data line 225. I do.

The search history display unit 208 includes a data line 225
Is displayed in a tree structure, and the search node ID of the designated node is output to the history management unit 204 via the data line 224.

The search history storage unit 205 stores a search history, and the document data storage unit 206 stores document data.

The data line 226 is connected to the browsing section 20.
In step 9, the browsing unit 209 transfers the permission / inhibition information of each document designated by the user from the browsing unit 209 to the search result display unit 207. Search result display unit 207
And the data lines 221 and 217 are
For transferring the immediately preceding search result, document availability information, and search node ID information at the time of re-search indicated by the search result display unit 207 from the search result display unit 207 to the search unit 203 and the history management unit 204 Data line.

FIG. 3 shows a functional configuration of the input analysis unit 202. The input analysis control unit 301 and the search command generation unit 30
2. Consists of a viewpoint extraction unit 303, a morphological analysis unit 304, and a syntax analysis unit 305.

The input analysis control unit 301 is activated from the above-mentioned input unit 201 by the control line 208. The input analysis control unit 301 activates the morphological analysis unit 304 via the control line 319, outputs each sentence to the morphological analysis unit 304 via the data line 312, and performs morphological analysis from the morphological analysis unit 304 via the data line 313. Enter the result. Also, the input analysis control unit 301 uses the control line 320 to
5 is started, and the morphological analysis result of each sentence is output to the syntax analysis unit 305 via the data line 314, and the data line 315
The syntax analysis result is input from the syntax analysis unit 305 via the. Further, the search command generation unit 302 is controlled by the control line 317.
And outputs the syntax analysis result to the search command generation unit 302 via the data line 308, and inputs the search command via the data line 309.

The search command generating unit 302 activates the viewpoint extracting unit 303 for each sentence by the control line 318, transfers the syntax analysis result via the data line 310, and
11, the viewpoint information extracted by the viewpoint extracting unit 303 is input. In addition, an independent word is extracted from each sentence, a search command is constructed, and a control line 309 is sent to the input analysis control unit 301.
Output via.

FIG. 4 shows the flow of processing of the input analysis control unit 301.

In this case, a plurality of natural language sentences may be input from the input unit. Therefore, morphological analysis and syntax analysis are performed on each input sentence (steps 401 and 402). The processing of the morphological analysis unit and the syntax analysis unit is processing already realized by machine translation or the like, and is not the gist of the present application. Therefore, the description here is omitted.

Then, based on the syntax analysis result analyzed corresponding to each sentence of the input sentence, processing for generating a search instruction is performed (step 403).

FIG. 5 shows the flow of processing of the search command generation unit 302.

In this case, the search command generation unit 302 extracts viewpoint information from the syntax analysis result of each sentence and also extracts independent words constituting the sentence (step 501,
502). Then, after performing viewpoint extraction and independent word extraction for all sentences constituting the input sentence, a search command is configured by merging the extracted independent words for each viewpoint (step 503).

Here, in extracting viewpoint information, FIG.
The extraction is performed based on a viewpoint extraction rule as shown in FIG. The viewpoint extraction rule has the following format.

<Collation part>===><viewpoint> In this case, a syntactic pattern is described in the collation part on the left side, and when collation is performed with this syntax pattern, the viewpoint described on the right side is replaced with the viewpoint to be processed. And

In the case where the viewpoint information cannot be extracted based on the viewpoint extraction rules, the search instruction generation unit 302 configures a search instruction with a preset default viewpoint. It should be noted that the default viewpoint can be set by inputting by the input means before the user executes the search command, in addition to the value set in advance.

For example, in the case of the rule on the first line in FIG. 6, the matching is performed with a subelement of a sentence including a syntax pattern “purpose”. In the case of the rule on the second line, the rule is checked against a subelement of a sentence including a syntax pattern "I found". In the illustrated example, the collating unit has the following format.

(<Predicate>, <modality>, <case structure>) For example, in the case of the rule on the first line, the predicate is “do”,
This indicates that the aspect is "continuous use" and the case structure is "(and purpose)". In the figure, “_” indicates that the pattern is to be collated with an arbitrary pattern.

Note that the format of the collating unit is not limited to the format of the exemplified rule according to the format of the syntax analysis result output by the syntax analysis.

In the independent word extraction, a noun portion is extracted from the result of the syntax analysis, and unnecessary words are removed using an unnecessary word dictionary (not shown).

FIG. 7 shows an example of a search command format. In this case, the search command is a pair of a plurality of keywords and a list of viewpoints. The illustrated example is a search command composed of an input sentence "A design tool based on a case has been developed for the purpose of computer design." The presence of “computer” and “design” in the noun controlled by “purpose” and the “case”, “design” and “tool” in the noun controlled by “developed” A search command as shown is configured.

FIG. 8 shows a functional configuration of the search unit 203.

In this case, the search unit 203 includes a search control unit 801 for controlling the entire search unit, a document data storage unit 206
An index scanning unit 802 that operates on a keyword index stored in the search unit and searches for a related document, a distance calculation unit 804 that calculates the similarity between the searched related document and the search command, and converts the similarity into a distance, a search result An adaptation unit 803 receives adaptability information of the document from the display unit and performs an adaptation process when re-searching is performed.

The processing performed by the search control unit 801 is classified into two types of processing: a search processing when a search command is input, and a re-search processing when document availability information is input from the search result display unit 207. Is done.

FIG. 9 shows the flow of the above-mentioned search processing.

In this case, when activated by the control line 230, this processing is executed. First, the keyword index scanning unit is activated by the control line 812 based on a search command input via the data line 214. On this occasion,
The search command is transferred to the index scanning unit 802 via the data line 813. In the index scanning unit 802,
The keyword index is scanned, a document ID including the keyword in the search command and viewpoint information are obtained as candidate documents (step 901), and this information is output to the search control unit 801 via the data line 807.

Then, for all the obtained candidate documents, the distance calculation unit 804 is activated by the control line 811 to calculate the distance with the search command (step 902). The distance calculation unit 804 inputs a keyword of a certain document and a search command from the search control unit 801 via the data line 808, and calculates a distance based on the following calculation formula. The result is output to the search control unit 801 via the data line 809.

[0049]

(Equation 1)

[0050]

(Equation 2) (Calculation formula for distance calculation) Dist (Q, Di) = | Q−Di | / M (3) Here, dij and qij are weights of keywords determined by analyzing documents and search commands. For example, the word i
Is used in the sentence of viewpoint j in the search instruction, qij is set to 1
And If not used, qij is set to 0. By using a synonym dictionary (not shown),
If you expand your keywords into synonyms,
It is also possible to assign a value of 1 or less and 0 or more to a synonym. Also, the position in the document structure where the keyword exists, for example, a title, a heading such as a chapter, a body,
It is also possible to assign a value between 0 and 1 according to a footnote or the like. These matrices represent feature vectors for search instructions and documents for each viewpoint.

| A | has the following meaning with respect to the matrix A.

[0052]

(Equation 3) Although the above expression expresses the city area distance, it can be changed to use another commonly used distance scale.

According to equation (3), the distance between the search command and each document is obtained for each viewpoint.

FIG. 10 shows a configuration example of the keyword index.

In this case, the keyword index is T
This is a format of the RIE structure. In the TRIE structure, the same character string of each keyword is shared and stored in order to reduce the storage capacity and the labor for searching.

The keyword index shown in FIG. 10 stores characters constituting each keyword and link information between the characters. For example, corresponding to the keyword “machine”, the link information of the character “machine” is “0030”
The link information “5” is stored. This link information indicates the address where the character “machine” is stored. Then, in the character “machine” stored at the address “00305”, a document including the keyword “machine” is “fi”.
le4 ”, that keyword is used in the sentence of the viewpoint“ conclusion ”, and link information about another keyword“ machine translation ”that includes“ machine ”as a partial character string of the preceding two characters, 00623 "is stored. By following this link information, documents using “machine translation” as a keyword are “file25” and “f25”.
ile21 ". In this embodiment, “f
"ile4" is a document ID for identifying a document.

As in the keywords “algorithm” and “arithmetic”,
A keyword including the same character string at the front is expressed by storing two or more addresses in link information of the character "calculation". For example, in this case, two addresses “00A15” and “00A16” are stored in the link information of the character “calculation”. These addresses represent the addresses where the characters "" and "" are stored, respectively.

Note that "0" in the link information indicates a separator for separating an address, a document ID, and the like. In addition, the first characters of all keywords are stored in a fixed continuous storage area in an order such as JIS code order.

FIG. 11 shows a format of the search result and an example of the search result.

In this case, a plurality of distances are assigned to each document ID. In the figure, the distance ij represents the distance of the viewpoint i to the document i.

The search result from the search unit 203 is output to the history management unit 204 via the data line 215 and the search result display unit 207 via the data line 216.
Is output to

FIG. 12 shows a second example of the search control unit 802.
Is shown.

The processing in this case is started from the search result display unit 207 by the control line 232. Search control unit 801
Receives the search result including the document availability information via the data line 217, activates the adaptation unit 803 via the control line 810, and reconstructs the search command (step 1201).
The calculation formula for reconstructing the search instruction in the adaptation unit 803 is shown below.

[0064]

(Equation 4) Q ′ thus obtained is used as a new search command.
After the recalculation of the search instruction, the same processing as described in FIG. 9 is performed (steps 1202 and 1203).

FIG. 13 shows a functional configuration of the history management unit 204. In this case, the history management unit 204 includes a history management control unit 1301 and a search node generation unit 1302.

FIG. 14 shows the flow of processing of the history management control unit 1301.

In this case, corresponding processing is performed according to the input message (step 1401). Now, the control line 23
When the generation of the search node ID is instructed from the input unit by 1 or from the search result display unit 207 by the control line 233, the generation of the search node is performed (step 14).
02). The generated search node is stored in the search history storage unit 20.
5 is stored.

FIG. 15 shows the format and an example of a search node.

In this case, the search node is the search node I
D, a parent node ID, a search instruction, and a search result. The search node ID stores an identifier for identifying the search node, and the parent node ID stores the ID of the search node storing the search command and the search result in the immediately preceding search. The history management unit 204 manages the search history based on the search node ID.

In generating a search node, the current search node is generated as a parent node. After generating the search node, the ID is output to the search result display unit 207 via the input unit 201 or the data line 220 via the data line 213.

When a pair of a search command and a search node ID is input from the input unit via the data line 212, and a pair of a search result and the search node ID are input from the search unit 201 via the data line 215, respectively , A search command, and a search result are stored (step 1403). Next, the search history display unit 20 is controlled by the control line 236 to update the search history display.
8 and the search history storage unit 2 via the data line 225.
05 is stored in the search history display unit 2
08 (step 1404).

Further, a change of the current search node is instructed from the search history display unit 208 by the control line 237.
In this case, the search history display unit 20 is connected via the data line 224.
In step 8, a search node designated by a mouse or the like is input, and the process in step 1405 is performed. In addition, while updating the search history display, the control line 238 instructs the search result display unit to update the display (step 140).
6, 1407).

FIG. 16 shows the flow of processing of the search history display unit 208.

In this case, corresponding processing is performed according to the input message (step 1601). First, the display window of the search history display unit 208 is instructed to be generated by the control line 235. In this case, step 1605
Is performed. If the control line 235 instructs the deletion of the search history display window, step 16
Step 06 is performed.

When a new history display is instructed from the history management unit 204 via the control line 236, the search history is input via the data line 225 (step 1602), and the search history is displayed in the search history display window (step 1602). 16
03). FIG. 17 shows an example of the displayed search history display.

In this case, as shown in FIG. 17A, the search nodes are displayed in a tree structure, in which black dots represent search nodes, and nodes connected by straight lines have a parent-child relationship. Also, a node surrounded by a white circle in FIG. 17A indicates that it is the current search node. In displaying the current search node, a display (blink display, reverse display, highlighted color display, etc.) for highlighting the difference from the other nodes is performed.

The search history display unit 208 receives an input from the user for an instruction to change the current search node and an instruction to generate a new search result display window. As an input method of the user when changing the current search node, an arbitrary search node displayed in the window is input by selecting it with a mouse or the like.

When an instruction to change the current search node is issued, step 1604 is performed. That is, the processing (steps 1405, 1406, and 1407) described with reference to FIG. 14 is instructed to the history management control unit 1301. By changing the current search node, a display different from that shown in FIG. 17A is obtained, as shown in FIG.

On the other hand, when an input from the user is received for generating a new search result display window, the control line 239 instructs the search result display unit to generate and display a new search result display window.

FIG. 18 shows the flow of processing in the search result display unit 207. The search result display unit 207 displays a search result input from the search unit 203 via the data line 216, and displays a search result in the history information input from the history management unit 204 via the data line 220. And input of the document availability information from the browse unit 209 via the data line 226, generation of a search result display window from the input unit by the control line 239, and generation of a search result display window by the control line 229. Do.

Further, by accepting the user's input in the search result display unit 207, the displayed viewpoint can be changed, the browse unit 209 can be started, and the search unit can be started based on the set document availability information. Do.

Input unit 201 or search history display unit 2
If the instruction to generate a search result display window is issued from 08, the processing of step 1802 is performed. The instruction from the search history display unit 208 corresponds to Step 1 shown in FIG.
607.

Further, when the search result is input from the search unit 203 via the data line 216 or when the search result in the history information is input from the history management unit 204 via the data line 220, The processing of 1803 is performed. In displaying the search result, multi-dimensional display is performed with the search command as the origin, based on the distance between the search command and the candidate document, with each viewpoint stored in the search result as an axis.

FIG. 19 shows an example of a search result display. FIG. 14 shows a display example of the search result shown in FIG. In this case, by displaying the distance between the document and the search command obtained for each viewpoint as coordinates, three-dimensional display is possible.

In FIG. 19, one of the axes of the viewpoint is “other”. When the viewpoint extracted from the retrieval statement exceeds a predetermined type, the space can be reduced and displayed by mapping the coordinates of a plurality of axes onto one axis. The “other” axis expresses the axis in the case where the space is reduced and displayed. In order to calculate the coordinates for the “other” axis, for example, a weighted average may be obtained by the following calculation formula.

[0086]

(Equation 5) Here, ri represents the distance of the degenerate axis i, and vi represents the weight for it. Display is performed with r as the coordinate of the axis “other”.

The above equation can be modified as follows.

[0088]

(Equation 6) Instead of displaying the character string “Others” near the axis, a list of degenerated axes can be displayed as it is (eg, “background + method”).

For each axis in the three-dimensional display, a viewpoint corresponding to the axis is displayed.

Although FIG. 19 shows an example of three-dimensional display, it is easy to change to two-dimensional or one-dimensional display depending on the number of viewpoint axes.

When the user inputs information indicating whether or not a document is being browsed from the browse unit 209, the data line 226 is displayed.
The information is input via the, and the processing of step 1804 is performed. The format of the input / output information of the input document is shown below.

<Document ID>;<PermissionInformation>. The permission information takes a value of 0 or 1, where 0 is inappropriate as a document to be searched and 1 is appropriate as a document to be searched. It means there is.

Next, the search result display unit 207 changes the display of the document corresponding to the input document ID. FIG.
(C) shows this state. Documents determined to be inappropriate are indicated by white circles, documents determined to be appropriate are indicated by black circles, and the propriety information of each document is displayed as color information.

The document availability information is stored as shown in FIG. 20 until a new search result is input.

[0095] The search result display unit 207 is operated by a user using a mouse or the like on the search result display window.
The browse unit 209 is activated.

If a point representing an individual document is selected on the search result display window, step 1806 is executed.
Is performed, and the browsing unit 209 is controlled by the control line 234.
Start An example of this display is shown in FIG. The point corresponding to the document currently being browsed is highlighted by blinking or changing the color.

When "re-search" is selected on the search result display window, the process of step 1807 is performed, and the search unit 203 is started. At this time, the search result including the information on the propriety of the document set up to that time is displayed on the data line 21.
7 to the search unit 203.

If "viewpoint" is selected on the search result display window, steps 1808 and 18
09 is performed. In the viewpoint change, which viewpoint is assigned to which axis is changed based on a user's input. Then, by displaying the search result again, the search result is displayed with the viewpoint changed.

If "generate" is selected on the search result display window, the process of step 1802 is performed to generate another search result display window. A separate window is created and displays the same search results as the original window. By selecting “viewpoint” for each window, search results can be displayed with different viewpoints as axes. However, since the search result and the document propriety information are shared, the display of the document propriety information is displayed equally.

When "delete" is selected on the search result display window, the search result display window is deleted (step 1810).

Next, FIG. 21 shows the flow of processing of the browse unit 209. The browse unit 209 is activated from the search result display unit 207. Then, input from the user is received, and the contents of the document to be displayed are displayed.

In the process of step 1806 shown in FIG. 18, when the search unit 203 is activated by the control line 234, the process of step 2102 is performed, and a window for displaying the browse unit is generated. At this time, data line 2
The document ID is input from the search result display unit 207 via the interface 27. Next, on the window, the document data to be displayed is read from the document data storage unit 206 via the data line 219, and the abstract sentence is displayed (the original sentence can be changed to be displayed). is there).

FIG. 22 shows an example of the contents of a document stored in the document data storage unit 206. The document data storage unit 206 stores information such as an original sentence, an abstract sentence, a keyword list, a viewpoint list, and a document structure for each document.

FIG. 23 shows an example of the display. In the process of step 2103, the stored abstract is displayed as shown in FIG.

When "viewpoint" is selected on the display window of the browse unit 209, the processing from step 2104 is performed. Thereby, as shown in FIG. 23C, a list of viewpoints stored in the document data to be displayed is presented as a menu. When the user selects a viewpoint from the displayed menu, a portion of the original text corresponding to the viewpoint is displayed. FIG.
(E) shows this state.

When "important word" is selected on the display window of the browsing unit 209, the processing from step 2108 is performed. As a result, as shown in the vinegar 23 (b), a list of keywords stored in the document data to be displayed is presented as a menu.

Next, when the user selects a certain keyword from the options, a list of viewpoints related to the keyword is presented as shown in FIG. When the user selects a viewpoint from the displayed menu, a portion of the original text corresponding to the viewpoint is displayed. FIG. 23E illustrates this state.

Therefore, in this way, the distance between the search command and the document can be obtained for each of a plurality of viewpoints, and the result can be displayed in a multidimensional space. In addition, because of the multidimensional display, it is possible for the user to understand the tendency of each document being similar to the search command from a list.

Further, it is possible to have the search result display before the re-search in the history. That is, by storing the search result and the document availability information as history information, it is possible to reproduce the search result display before performing the re-search.

In the first embodiment, as shown in FIG. 10, the keyword index structures the word corresponding to the keyword in the TRIE structure, and stores the word in the viewpoint information and the pointer information to the document before the word. . It is also possible to have a plurality of such keyword indices, and to configure each of these keyword indices to express keyword information from another viewpoint. In this case, the information stored for each word in the TREI structure is only a pointer to the document.

In addition to the TREI structure, as a means for performing dictionary lookup at high speed, a method using a hash function is generally known. However, such a method can be used to construct a keyword index. is there.

Further, in the first embodiment, the viewpoint information is extracted from the input retrieval command sentence, and if it cannot be extracted, the default viewpoint is used. However, for example, as shown in FIG. 28, by configuring an input window for a search statement and providing a button for setting a viewpoint in the input window, the viewpoint can be set in a menu format as needed. In this case, the search command is configured using the viewpoint set immediately before the search command statement is input and ended (input of the statement and input of a new line at the end).

At the right end of the window shown in FIG. 28, information on what viewpoint has been set is displayed. In other words, it indicates that an instruction has been issued to perform a search that concludes "development of an expert system" for the purpose of "sophistication of knowledge".

(Second Embodiment) Next, a second embodiment of the present invention will be described. The second embodiment is different from the first embodiment in that the processing content of the search result display unit 207 is changed.

FIGS. 24 and 25 show the flow of processing of the search result display unit 207 for the second embodiment. The components other than the search result display unit 207 are the same as those in the first embodiment, and a description thereof will be omitted.

Processing started from other components includes generation of a search result display window (step 240).
3), display of search results (step 2404), setting of document availability information (steps 2405, 2406), and the like. Since these processes are the same as those in the first embodiment, description thereof will be omitted. In addition, as a process for mouse input, the search result display window is deleted (step 240).
8), generation of a search result display window (step 24)
10), re-search based on document availability information (step 24)
12), change of viewpoint (steps 2414 and 2415),
The activation of the browsing unit (step 2417) is the same as that of the first embodiment, and a description thereof will not be repeated.

When the user selects an axis representing each viewpoint with the mouse, the axis is rotated in accordance with the movement of the mouse by the processing of step 2419, and the axis is enlarged and reduced. The search result is displayed again based on the arranged axis (step 2420). This situation is shown in FIG.
(B) and (c).

If the user selects “enlarge”, process 2
After specifying a range by mouse input at 422, an enlarged display is performed with the specified range as a whole. That is, after the distance is converted by the following formula, display is performed based on the distance.

[0119]

(Equation 7) FIG. 26D shows a display state in which the range is specified.

When the user selects “reduction”, a range is specified by the processing in step 2425, and then reduced display is performed so that the specified range becomes the whole. That is, after the distance is converted by the following formula, display is performed based on the distance.

[0121]

(Equation 8) When the user selects “appropriateness”, the range is specified by the process of step 2428, and then the rejection information is inverted and the re-display is performed for the document in the specified range.

As for the process of specifying the range, it is possible to change the configuration so that the range is specified first, and then the instruction indicating the content of the process is input.

It is possible to express the perspective of the origin by overwriting the point or title of the document located earlier on the point or title of the document located later. Needless to say, various methods for displaying a three-dimensional object in computer graphics, such as color change or shading of the distance between the title and the point representing the document and the origin and the origin, can be used.

(Third Embodiment) In the first embodiment, the search history is displayed in a tree structure, but in the third embodiment,
In displaying the search history, a reduced display of the search result display corresponding to the search node is performed.

An example of the display is shown in FIG. By displaying the search results in a reduced size, the user can easily list past searches.

For the reduced display of the search result, the method described in the second embodiment may be used. (Fourth Embodiment) FIG. 29 shows the functional configuration of the fourth embodiment. In this case, FIG. 29 replaces the leading numeral “2” of the reference numeral assigned to each part in FIG. 2 with “29”, and attaches the same reference numeral to the same part in the last two digits of each reference numeral. Since each part is the same as that of FIG. 2, the description is omitted here.

The functional configuration of the input analysis unit 2902 shown in FIG. 30 and the flow of processing of the input analysis control unit 301 shown in FIG. 31 are also described with reference numerals “2” and “3” assigned to each unit in FIG. Are replaced with "29" and "30", and the leading numeral "4" of the reference numeral assigned to each part in FIG. 4 is replaced with "31". Since each part is the same as that in FIGS. 3 and 4, the description is omitted here.

FIG. 32 shows the flow of processing of the search command generation unit 3002.

In this case, the search instruction generation unit 3002
The viewpoint information is extracted from the syntax analysis result of each sentence, and the independent words constituting the sentence are extracted (step 320).
1, 3202). Then, after performing viewpoint extraction and independent word extraction for all sentences constituting the input sentence,
By merging the extracted independent words for each viewpoint,
A search command is configured (step 3203). Further, a window for displaying the viewpoint extraction result is generated (step 320).
4) The viewpoint extraction result is displayed in the window (step 3205).

Here, in the extraction of the viewpoint information, FIG.
The extraction is performed based on a viewpoint extraction rule as shown in FIG. The viewpoint extraction rule has the following format.

<Matching Unit>===><Viewpoint> In this case, a syntax pattern is described in the matching unit on the left side, and when matching with this syntax pattern, the viewpoint described on the right side is replaced with the viewpoint to be processed. And

If the viewpoint command cannot be extracted based on the viewpoint extraction rules, the search command generator 3002 configures a search command with a preset default viewpoint. It should be noted that the default viewpoint can be set by inputting by the input means before the user executes the search command, in addition to the value set in advance.

For example, in the case of the rule on the first line in FIG.
Match the sentence sub-element containing the syntactic pattern “purpose”. In addition, in the case of the rule on the second line, "I understood"
Is matched with the subelement of the sentence containing the syntax pattern In the illustrated example, the collating unit has the following format.

(<Predicate>, <modality>, <case structure>) For example, in the case of the rule on the first line, the predicate is “do”,
This indicates that the aspect is "continuous use" and the case structure is "(and purpose)". In the figure, “_” indicates that the pattern is to be collated with an arbitrary pattern.

Note that the format of this collating unit is not limited to the format of the exemplified rule according to the format of the syntax analysis result output by the syntax analysis.

In the independent word extraction, a noun portion is extracted from the result of the syntax analysis, and unnecessary words are removed using an unnecessary word dictionary (not shown).

FIG. 34 shows an example of the format of a search command. In this case, the search command is a pair of a plurality of keywords and a list of viewpoints. The illustrated example shows that "a case-based design tool has been developed for the purpose of computer design."
This is a search command composed of the following input sentences. "Computer" and "design" are nouns governed by "purpose"
Nouns dominated by “developed”
The presence of “design” and “tool” forms a search command as shown.

FIG. 35 shows a display example of the viewpoint extraction result. In this case, FIG. 35 shows an example in which the search command shown in FIG. 34 is displayed.

FIG. 36 illustrates the functional configuration of the search unit 2903. In this case, FIG. 36 replaces the leading numerals “2” and “8” of the reference numerals assigned to the respective parts in FIG. Since each part is the same as that in FIG. 8, the description is omitted here.

FIG. 37 shows the flow of a search process in such a search unit 2903.

In this case, this process is performed when activated by the control line 2930. First, the keyword index scanning unit is activated by the control line 3612 based on a search command input via the data line 2914. At this time, the search command is transferred to the index scanning unit 3602 via the data line 3613. Index scanning unit 36
In step 02, the keyword index is scanned, a document ID including the keyword in the search command and viewpoint information are obtained (step 3901), and this information is output to the search control unit via the data line 3607.

Then, for all the obtained candidate documents, the distance calculation unit 3604 is activated by the control line 3611, and the distance to the search command is calculated (step 370).
2). The distance calculation unit 3603 inputs a keyword, viewpoint information, and a search command of a certain document from the search control unit 3601 via the data line 3608, and the above-described (1) to (3).
The distance calculation is performed based on the calculation formula (4). This result is output to search control section 3601 via data line 3609.

As described above, the distance between the search command and each document is obtained for each viewpoint by the equation (3).

Next, FIG. 38 shows a configuration example of the keyword index. Since the keyword index in this case is the same as the configuration in FIG. 10 described above, the description here is omitted.

FIG. 39 shows the format of the search result and an example of the search result.

In this case, a plurality of distances are assigned to each document ID. In the figure, the distance ij represents the distance of the document i with respect to the viewpoint j. Then, the search unit 29
03 is output to the history management unit 2904 via the data line 2915 and the data line 2916
Is output to the search result display unit 2907 via the.

FIG. 40 shows the second processing in search control section 3602.

The processing in this case is started from the search result display unit 2907 by the control line 2932. Search control unit 3
The control unit 601 receives a search result including information on whether or not a document is input via the data line 2917, activates the adaptation unit 3603 via the control line 3610, and reconstructs a search command (step 4001). The calculation formula for reconstructing the search instruction in the adaptation unit 3603 is as the above-described formula (5).

[0149] The thus obtained Q 'is used as a new search instruction. And after recalculating the search instruction,
The processing is similar to that described in FIG. 37 (step 400).
2,4003).

FIG. 41 shows the functional configuration of the history management unit 2904. In this case, the history management unit 2904 includes a history management control unit 4101 and a search node generation unit 4102.

FIG. 42 shows the flow of processing of the history management control unit 4101.

In this case, corresponding processing is performed according to the input message (step 4201). Now, control line 29
When an instruction to generate a search node ID is issued from the input unit 31 or from the search result display unit 2907 through the control line 2933, the search node is generated (step 4202). The generated search node is stored in the search history storage unit 2905.

FIG. 43 shows the format of a search node and an example.

In this case, the search node is the search node I
D, parent node ID, search date and time, search input sentence, search command,
It consists of the number of search cases, the search result, the number of OK documents, the number of OK documents, the number of NG documents, and the NG documents. The search node ID stores an identifier for identifying the search node, and the parent node ID stores the ID of the search node storing the search command and the search result in the immediately preceding search. The history management unit 2904 manages the search history based on the search node ID. In the search date and time, the date of the year in which the search was performed and the time in seconds are stored. The search input sentence, the search command, the number of search cases, and the search result store information when the search is performed. Number of OK documents, OK
In the document, the number of NG documents, and the NG document, the browse unit 29
In step 09, information indicating whether the user can set the document is stored.

After the search node is generated, its ID is
The data is output to the search result display unit 2907 via the input unit 2901 via the data line 2913 or via the data line 2920.

Input unit 2901 via data line 2912
, And a pair of a search result and a search node ID from the search unit 2903 via the data line 2915, respectively, the search command and the search result are stored. (Step 420
3). Next, a control line 2936 is used for updating the search history display.
Is instructed to the search history display unit 2908 by the
The respective search nodes stored in the search history storage unit 2905 are transferred to the search history display unit 2908 via the server 25 (step 4204).

In addition, control line 2937 instructs change of the current search node from search history display portion 2908. In this case, a search node input and instructed by a mouse or the like is input to the search history display unit 2908 via the data line 2924, and the process of step 4205 is performed. Also,
The search history display is updated, and the control line 2938 instructs the search result display unit to update the display (steps 4206 and 4207).

FIG. 44 shows the flow of processing of the search history display unit 2908.

Also in this case, corresponding processing is performed according to the input message (step 4401).

First, the display window of the search history display unit 2908 is instructed to be generated by the control line 2935. In this case, the process of step 4405 is performed.
If the control line 2935 gives an instruction to delete the search history display window, the process of step 4406 is performed.

The history management unit 2904 is controlled by the control line 2936.
Is instructed to display a new history by the data line 292
The user inputs a search history via the terminal 5 (step 4402), and displays the search history in a search history display window (step 4403). An example of the displayed search history display is the 45th
Is shown in

In this case, as shown in FIG. 45 (a), the search history is displayed in a tree structure, and in this display, black circles represent each search node, and the nodes connected by a straight line have a parent-child relationship. is there. In the vicinity of the black circle, the search date and the number of search nodes of the search node represented by the black circle are displayed. A node surrounded by a white circle indicates a current search node. In displaying the current search node, a display (blink display, reverse display, highlighted color display, etc.) for highlighting the difference from the other nodes is performed.

The search history display unit 2908 receives an input from the user for an instruction to change the current search node and an instruction to generate a new search result display window. To change the current search node, the user selects an arbitrary search node displayed in the window with a mouse or the like.

When an instruction to change the current search node is issued, step 4404 is performed. That is, it instructs the history management control unit 4101 to perform the processing described in FIG. 42 (steps 4205, 4206, and 4207). By changing the current search node, a display different from that shown in FIG. 45A is obtained, as shown in FIG.

On the other hand, when an input from the user is received for generating a new search result display window, the control line 2939 instructs the search result display unit 2907 to generate and display a new search result display window ( Step 441607). In addition, a search node information display window is generated (step 441610), the search node information of the changed current search node is read (step 4408), and the read search node information is displayed in the search node information display window. (Step 4409).

FIG. 46 shows a display example of the search node information on the search node information display window. In this case, the display content in FIG. 46 shows a case where the example of the search node shown in FIG. 43 is displayed. If the user specifies an OK document name or an NG document name with a mouse or the like, an instruction is given to generate a browse section display window and display the specified document (step 4412).

FIG. 47 shows the flow of processing of the search result display unit 2907. In the search result display unit 2907, display of search results input from the search unit via the data line 2916, display of search results in history information input via the data line 2920 from the history management unit 2904, and , Input of document availability information via the data line 2926 from the browse unit 2909, generation of a search result display window from the input unit by the control line 2939, and generation of a search result display window by the control line 2929. .

Further, by accepting a user's input in the search result display unit 2907, the displayed viewpoint can be changed, the browse unit 2909 can be activated, and the search unit can be activated based on the set document availability information. Do.

If an instruction to generate a search result display window is issued from input unit 2901 or search history display unit 2908, step 4202 is performed. The instruction from the search history display unit 2908 corresponds to Step 4 shown in FIG.
407 corresponds to the process.

The search unit 2903 sends the data line 291
6 is performed, or the search result in the history information is input from the history management unit 2904 via the data line 2920, whereby the step processing 4703 is performed. In displaying the search result, multi-dimensional display is performed with the search command as the origin based on the distance between the search command and the candidate document with each viewpoint stored in the search result as an axis.

If the number of search cases does not exceed the number of clustered documents set in advance, the document of the search result is displayed in multi-dimension (step 471803). If the number of search cases exceeds the number of clustered documents set in advance, clustering is performed by combining a plurality of documents within a certain distance into one cluster (step 471).
2), display in a search result display window (step 4713).

FIG. 48 shows the flow of the clustering process.

First, the documents of the retrieval result are clustered such that the number of documents in one cluster is equal to or less than a preset number of documents in the cluster (step 480).
1). Next, if the number of generated clusters exceeds the preset number of clusters, the cluster having the largest number of documents in the cluster is further divided into clusters until the number of clusters becomes equal to or less than the number of clusters (step 480).
3).

FIG. 49 shows the flow of the clustering process.
In this case, in the process of clustering in steps 4901 to 4903, cluster / document correspondence data shown in FIG. 50 is generated so that correspondence between clusters and documents can be obtained.

FIG. 51 shows a display example of a document cluster on the search result display window. In this case, the distance between the cluster generated for each viewpoint and the search command is displayed as coordinates, thereby displaying three-dimensionally. The viewpoint is displayed in association with each axis in the three-dimensional display. Clusters are represented by dots. Also, the title of an arbitrary document in the document cluster represented by the point can be displayed near the displayed point. The point at which the cluster is displayed is calculated from the coordinate position of the document included in the cluster. For example, the average of the coordinate positions of the documents in the cluster is calculated.

The points at which the clusters are displayed are displayed by changing the size, shading, etc. of the points in accordance with the number of documents included in the cluster. Alternatively, the number of documents included in the cluster can be displayed in or near the point.

[0177] Further, the point indicating the cluster located earlier is overwritten on the point indicating the cluster located later, thereby expressing the perspective with respect to the origin. In addition, various methods for displaying a three-dimensional object in computer graphics, such as color change and shading of the distance between the point displaying the cluster and the origin, can be used. Alternatively, numbers may be assigned in ascending order of the distance to the origin, and the numbers may be displayed in or near the points.

When the user designates one or more clusters with the mouse or the like on the search result display window and selects “expand”, a new search result display window is generated (step 4702), and the designated window is designated. The documents in the cluster are displayed in a multi-dimensional manner in the window (step 470).
3). At this time, an enlarged display is performed using the coordinate range where the document in the designated cluster exists as the entire coordinate. Also,
When "cluster" is selected on the search result display window, the set number of clustered documents, the number of documents in a cluster, the number of clusters, and the like are displayed (step 471).
4, 4715). The user can change these values. Further, when "redisplay" is selected, processing of displaying the search result is performed according to the change of the numerical value.

If the user cannot properly designate a point for displaying a cluster on the search result display window, the coordinate range in which a cluster group located at a certain distance from the designated position exists is set to the coordinate range. It can be enlarged and displayed as a whole. Alternatively, it is also possible to perform processing as if the cluster closest to the specified position was specified. Further, when a cluster is designated on the search result display window, it is also possible to change so that documents in the designated cluster are displayed one by one from the browse unit in a predetermined order. Alternatively, the titles of the documents in the document cluster, which are displayed near the point where the cluster is displayed, can be displayed in a predetermined order.

Alternatively, the documents in a certain number of clusters may be displayed in the generated search result display window in advance in the order of proximity to the origin without the user specifying a cluster. Alternatively, for a certain number of clusters in the descending order of the number of included documents, the documents in the clusters may be displayed on the search result display window. In these cases, it is possible to display a correspondence between a point expressing a cluster and a search result display window displaying documents in the cluster.

FIG. 52 shows an example of a search result display in which documents in the cluster are displayed when a cluster is specified on the search result display window and “expansion” is selected.

In this case, FIG. 52A shows an example in which the search results shown in FIG. 39 are displayed. The distance between the document and the search command obtained for each viewpoint is displayed as coordinates and displayed three-dimensionally. The viewpoint is displayed in association with each axis in the three-dimensional display. The point or title representing the document located earlier is overwritten on the point or title representing the document located later to express the perspective with respect to the origin. In addition, various methods for displaying a three-dimensional object in computer graphics, such as color change or shading of a point representing a document or the distance between a title and an origin, can be used. Alternatively, numbers are assigned in ascending order of the distance to the origin, and the numbers are
It can also appear near the point or before the title.
In FIG. 52, the number after “the number of items in the cluster” is
Indicates the number of documents in the specified cluster, that is, the number of currently displayed documents.

Note that, even when the search result is not clustered because the number of search cases does not exceed the number of clustered documents, the document of the search result is similarly displayed in a multidimensional manner (step 4703). However, “the number of items in the cluster” is not displayed.

Here, in FIGS. 51 and 52, one of the axes of the viewpoint is “other”. When the number of viewpoints extracted from the retrieval statement exceeds a predetermined number, the space can be reduced and displayed by mapping the coordinates of a plurality of axes onto one axis. “Other”
Axis represents an axis in a case where the space is reduced and displayed. In order to calculate the coordinates with respect to the “other” axis, for example, a weighted average may be obtained by the above-described calculation formula (6).

Here, since the above equation can be modified as in the above equation (7), these equations are
In general terms, it can be expressed as in equation (8) above.

It is also possible to display the list of degenerated axes as they are, instead of displaying them in the vicinity of the axis with the character string “Others” (for example, “background + method”).

In FIGS. 51 and 52, the scales on the axis of the viewpoint are equally spaced. It can also be displayed as a logarithmic scale. It is also possible to allow the user to select those settings.

In FIG. 51, the range of the axis of the viewpoint is from 0 to 100.
In FIG. 52, the range of the viewpoint axis corresponds to the entire space divided when the cluster including the document illustrated in FIG. 52 is generated. The range of axes may be the location of the cluster or document that is at the greatest distance for each axis.

FIG. 51 and FIG. 52 show examples of three-dimensional display, but it is easy to change to two-dimensional or one-dimensional display depending on the number of viewpoint axes.

A search result display window (FIG. 51)
Alternatively, if “viewpoint” is selected on (FIG. 52), steps 4708 and 4709 are performed. In the viewpoint change, which viewpoint is assigned to which axis is changed based on a user's input. Then, by displaying the search result again, the search result is displayed with the viewpoint changed.

If "generate" is selected on the search result display window (FIG. 51 or 52), another search result display window is newly generated.
Is performed. A separate window is created and displays the same search results as the original window. By selecting “viewpoint” for each window, search results can be displayed with different viewpoints as axes. However,
Since the search result and the information on whether or not the document is input in the browsing unit are shared, the display of the information on whether or not the document is acceptable is equally displayed.

When "delete" is selected on the search result display window (FIG. 51 or 52), the search result display window is deleted (step 4710).

When a document is displayed on the search result display window, the search result display unit 2907 activates the browse unit 2909 by designating the document with the user's mouse or the like. If a point representing an individual document is selected on the search result display window, the process of step 4906 is performed, and the browse unit 290 is controlled by the control line 2934.
9 is started. An example of this display is shown in FIG.
In this case, the point corresponding to the document currently being browsed is
Perform highlighting, such as blinking or changing colors.

When the user inputs information about whether or not the document being browsed from the browse unit 2909, the data line 29
The information is input via the interface 26, and the process of step 4704 is performed. In this case, the format of the input / output information of the input document is shown below.

<Document ID>;<Possibilityinformation>. The propriety information takes a value of 0 or 1, where 0 is inappropriate as a document to be searched and 1 is a document to be searched. As appropriate.

Next, the search result display portion 2907 changes the display of the document corresponding to the input document ID. FIG.
This situation is shown in FIG. Here, documents determined to be inappropriate are indicated by white circles, documents determined to be appropriate are indicated by black circles, and the propriety information of each document is displayed as color information.

The document availability information is stored until the search is completed.
It is stored as shown in FIG. When the search is completed, the stored document availability information is transferred to the search history management unit and stored in the current search node (step 42).
09).

When "re-search" is selected on the search result display window, step 4707 is performed, and the search unit is started. At this time, the search result including the information about whether the document has been set up to that point is output to the search unit 2903 via the data line 2917.

When "Alignment" is selected on the search result display window, the user activates the browse unit for the specified document and specifies the browse unit display window generated by the browse unit. (Step 4711).

FIG. 54 shows an example of a menu displayed when the user selects “arrange” on the search result display window.

In this case, “distance from origin” indicates the order of closest to the origin in the coordinates, “document creation date” indicates the latest document creation date, “author's syllabary” indicates the author's syllabary order, In "", the browsing section display windows are arranged in ascending order of the date and time when the document was displayed.

FIG. 55 (a) shows a browse part display window before the user selects “arrange” on the search result display window, and FIG. 55 (b) shows the search result display window. The window for browsing part display after selecting "alignment" above is illustrated.

In FIG. 55, corresponding to a specific order,
It is possible to change the color, shading, etc., such as the frame of each window and the background in the window. For example, a window displaying a document close to the origin in coordinates, a window displaying a document with a new document creation date, a window displaying a document starting with the author's syllabary order, or a document displaying a document with a new date and time. Is displayed, the color of the frame of the window or the background in the window can be made darker for the window.

FIG. 56 shows the flow of processing of the browsing unit 2909.

[0205] The browsing unit 2909 includes the search result display unit 2
907 is started. Then, input from the user is received, and the contents of the document to be displayed are displayed.

In step 4706 in FIG. 47, when the search unit is activated by the control line 2934, the process of step 5602 is performed, and a window for displaying the browse unit is generated. At this time, a document ID is input from the search result display unit 2907 via the data line 2927. Next, the document data to be displayed is read from the document data storage unit 2906 via the data line 2919 on the window, and the abstract sentence is displayed (the original sentence can be changed to be displayed). is there).

FIG. 57 shows an example of the contents of a document stored in the document data storage unit 2906. In this case, the document data storage unit 2906 stores information such as original text, abstract text, keyword list, viewpoint list, document structure, and time information for each document.

FIG. 58 shows a display example. In this case, in step 5603 of FIG. 56, the stored abstract is displayed as shown in FIG. 58 (a).

[0209] If "OK" is selected on the display window of the browse unit 2909, the propriety information of the displayed document is set to 1, and if "NG" is selected, the propriety information of the displayed document is set. The value is set to 0 and transferred to the search result display section (step 5615).

If "viewpoint" is selected on the display window of the browsing unit 2909, the processing from step 5604 is performed. As a result, as shown in FIG. 58C, a list of viewpoints stored in the document data to be displayed is presented as a menu. When the user selects a viewpoint from the displayed menu, a portion of the original text corresponding to the viewpoint is displayed. FIG. 58 (e)
FIG.

If "important word" is selected on the display window of the browse unit 2909, the processing from step 5608 is performed. Thus, as shown in FIG. 58B, a list of keywords stored in the document data to be displayed is presented as a menu.

Next, when the user selects a certain keyword from the options, a list of viewpoints related to the keyword is presented as shown in FIG. 58 (d). When the user selects a viewpoint from the displayed menu, a portion of the original text corresponding to the viewpoint is displayed. FIG. 58E shows this state.

Therefore, even in this case, the distance between the search command and the document can be obtained for each of a plurality of viewpoints, and the result can be displayed in a multidimensional space. Because of the multi-dimensional display, it is possible for the user to understand, from a list, the tendency of each document being similar to the search command.

[0214] The search result before the re-search is performed can be stored in the history. That is, by storing search results and document availability information as history information,
The display of the search result before the re-search can be reproduced.

In the fourth embodiment, as shown in FIG. 38, the keyword index structures the word corresponding to the keyword in the TRIE structure, and stores the viewpoint information and the pointer information to the document at the end of the word. . On the other hand, a keyword index may be provided for each viewpoint, and keyword information may be expressed in each of these keyword indexes. In this case, the information stored for each word in the TREI structure is only a pointer to the document.

In addition to the TREI structure, a method using a hash function is generally known as a means for performing dictionary lookup at high speed. However, it is also possible to construct a keyword index using such a method. is there.

(Fifth Embodiment) In the fourth embodiment, cluster display is performed in the search result display window (FIG. 5).
1) When the user designates a point representing a cluster, the documents in the designated cluster are displayed in a multidimensional manner.
When a document is displayed in the search result display window (FIG. 52), when the user specifies a point expressing the document or a title, the content of the specified document is displayed.

On the other hand, in the fifth embodiment, when the user designates a point representing a cluster, the propriety information of the document in the designated cluster can be set to 1 or 0. Further, when the user designates a point representing a document or a title, the propriety information of the designated document can be set to 1 or 0. These processes are performed in the search result display unit (2907 in FIG. 29).

FIG. 59 shows a display when the user designates “document” on the window when a cluster is displayed in the search result display window.

In this case, when the user designates “document” after designating one or more clusters and selects “expansion”, as in the fourth embodiment, the number of documents in the designated cluster is increased. The dimension is displayed (step 4703 in FIG. 47). Here, when “OK” is selected, the propriety information of all documents in the designated cluster is set to 1. When “NG” is selected, the propriety information of all documents in the specified cluster is set to 0. At this time, in the search result display window, points representing clusters selected for “expand” or “OK” are displayed as black circles, points representing clusters selected for “NG” are displayed as white circles, The selection result of the cluster is displayed as color information. Alternatively, it is also possible to display the result of selecting a cluster in shades of color. The document propriety information is transferred to the history management unit (step 4717 in FIG. 47).

FIG. 60 shows a display when a user designates “select” on the window when a document is displayed in the search result display window. In this case, when the user designates "selection" after designating one or more documents and selects "display", the contents of the designated document are displayed in the browse section display window as in the fourth embodiment. (Step 4706 in FIG. 47). here,
When “OK” is selected, the validity information of the designated document is set to 1. When “NG” is selected, the validity information of the designated document is set to 0. At this time, in the search result display window, points and titles representing the document selected “display” or “OK” are displayed in black, and points and titles representing the document selected “NG” are displayed in white. And the selection result of the document is displayed as color information. Alternatively, it is also possible to display the result of selecting a cluster in shades of color. The document propriety information is transferred to the history management unit (step 4717 in FIG. 47).

Since the processing other than the above processing is the same as that of the fourth embodiment, the description is omitted here.

FIGS. 59 and 60 show examples of three-dimensional display, but they can be displayed one-dimensionally or two-dimensionally.
In addition, it is also possible that the user designates a specific range and designates a point or title included in the area by specifying a point representing a cluster, a point representing a document, or a title.

(Sixth Embodiment) In the first embodiment, multi-dimensional coordinates are obtained in the search result display window, and clusters are displayed as dots. In the fourth embodiment, however, the cluster is displayed at the coordinate position of the cluster. The number of included documents is displayed, and in a tabular form, the degree of similarity of each document from each viewpoint is displayed.

FIG. 61 shows a search result display window in which coordinates are two-dimensionally displayed and the number of documents included in the cluster is displayed at the coordinate position of the cluster. FIG.
Reference numeral 2 denotes a search result display window that displays the document name and the similarity of each document for each viewpoint.

(Seventh Embodiment) In the first embodiment, processing using a viewpoint was performed when searching for a document. However, in the seventh embodiment, the search is performed by checking all the indexes in the document data storage unit. The viewpoint of the document in the document data storage unit is displayed regardless of the processing. Thus, when there are a plurality of document data storage units, the viewpoints of the documents in each document data storage unit and the number of documents having each viewpoint can be displayed.
When a document in the document data storage unit is updated,
It is also possible to display the viewpoints of the documents in the document data storage unit and the number of documents having each viewpoint.

FIG. 63 shows a window displaying the name of the document data storage unit, the total number of documents and the number of documents including the displayed viewpoint for each document data storage unit.

(Eighth Embodiment) In the first embodiment, in the search processing, the similarity between a search command and a document is calculated for viewpoints such as the purpose, conclusion, and background extracted by the viewpoint extraction rules, and based on the similarity. In the eighth embodiment, numerical values such as a date, a price, and a computer speed are extracted by a numerical expression extraction rule, and the multidimensional display is performed based on the numerical values.

Here, the numerical expression extraction rule has the following format. <Collation unit>===><Type of numerical expression> The collation unit has the following format. (<Numerical expression>, <predicate>, <tense>) FIG. 64 shows an example of a numerical expression extraction rule.
In the numerical expression extraction rules shown in (1), the numerical values in the X, Y, and Z portions are extracted.

FIG. 65 shows a search result display window when the coordinate axes are the release date and price. The circle on the window represents a document such as a newspaper article relating to an IC memory or the like, including information on a release date and a price corresponding to the coordinate position.

(Ninth Embodiment) In the first embodiment, the user designates a point representing a cluster, a point representing a document, or a title with a mouse or the like. A range can be input or specified, and points and titles included in the range are enlarged and displayed as the whole coordinates.

FIG. 66 (a) shows a search result display window immediately after the search processing, and FIG. 66 (b) shows a search result display window after the user inputs or designates the axis range. Shows a window.

Note that not only the search result display window after the user has input or specified the axis range, but also both the window and the search result display window immediately after the search processing may be displayed.

(Tenth Embodiment) In the tenth embodiment, weighting is applied to each viewpoint extraction rule. In addition, the user can register viewpoint extraction rules and set weights.

Here, the viewpoint extraction rule has the following format.

<Collation Unit>===><viewpoint>, Weight FIG. 67 shows an example of weighted viewpoint extraction rules.

Next, the search command, document matrix expression, and calculation formula for distance calculation according to the present embodiment corresponding to the search command, document matrix expression, and calculation formula for distance calculation of the first embodiment will be described. Shown below.

[0238]

(Equation 9)

[0239]

(Equation 10)

[0240]

[Equation 11]

[0241]

(Equation 12) FIG. 68 illustrates an example of registration of a viewpoint extraction rule in the input unit.

In this case, FIG. 68A shows an example of a display when the user inputs a character string for registering a viewpoint extraction rule. FIG. 68B shows an example in which the input character string is analyzed by the input analysis unit and the analysis result is displayed. At this time, the weight is set to 1. FIG. 68 (c) shows that the user
This is an example in which a viewpoint is selected and a list of viewpoints is displayed as a menu. FIG. 68 (d) shows a case where the user selects a “purpose” in FIG. This is an example of a display in the case of correction, in which the user deletes “improvement” and sets the weight to 2. FIG.
In (d), when the user selects “confirmation”, the displayed viewpoint extraction rule is stored in the viewpoint extraction rule dictionary of the input analysis unit.

(Eleventh Embodiment) In the eleventh embodiment, the search history displayed in the search history display section window can be edited. The search results are compared and displayed.

FIG. 69 (a) shows a search history display window before editing the search history, and FIG. 69 (b)
9 shows a window for a search history display unit after the search history has been edited. FIG. 70 shows a display when the user selects “comparison” on the search history display section window and then specifies two search nodes. In FIG. 70,
The search nodes specified by the user are displayed as white squares and white circles. On the right side, information about the two search nodes is displayed. (Twelfth Embodiment) In a twelfth embodiment, document link information, which is information indicating a relationship between a plurality of selected documents, is set in a browse section display window, and the browse section sets this document link information. Displays the line to be represented. The format of the document link information is shown below. The link name is
It is a character string such as "reference" or "original text".

<Document ID No. 1>;<LinkName>;<Document ID No. 2> FIG. 71 shows an example of a browse section display window. In this case, when the user selects “link” on the browse section display window, a list of link names is presented as a menu. When the user selects one of them, the browse section displays the document ID of the document displayed in the window specified immediately before by the user and the document ID of the document displayed in the window. The document link information is set using the link name, and the document link information is displayed. FIG. 72 shows an example of a browse section display window on which document link information is displayed.

Further, in the twelfth embodiment, the personal information storage means is provided, and the user can access document link information, document availability information, search history, part of search history, search result, and specific part of search result. And so on can be stored in the personal information storage means. When storing such information in the personal information storage means, the user can create a document for a memo, add it, and store it. FIG. 73 shows an example of data in the personal information storage means.

(Thirteenth Embodiment) In the thirteenth embodiment, in the multidimensional display in the browse unit, the arrangement of axes and the setting of the viewpoint for each axis can be performed by a method of selecting from a menu. 74 and 75 show examples of the display of the search result display section window. FIG. 74 shows a display when the user selects "arrangement" on the search result display window. The user can change the display by selecting from the menu. In FIG. 74, when one-dimensional display is selected, one-dimensional display is performed, and when two-dimensional display is selected, two-dimensional display is performed. Similarly, in the search result display window, a change in the arrangement of viewpoints marked with A, B, and C can be selected from a menu.

FIG. 75 shows a display when the user designates a character string expressing the viewpoint on the search result display window. The user can change the viewpoint by selecting from the menu. it can. "Other" is the superposition of viewpoints other than the viewpoint set and set for the axis. Further, in the present embodiment, an adaptive learning unit is provided, and the arrangement of viewpoints and axes selected by the user is stored in the adaptive learning unit. In the next and subsequent search processing, display is performed with reference to the arrangement of viewpoints and axes stored in the adaptive learning means.

(Fourteenth Embodiment) In the fourteenth embodiment, the browse part display window generated by the browse part by performing one search is displayed as it is even when the next search is performed, and the search part is displayed by the search. The window display method and the like are changed depending on the generated browse section display window. For example,
In accordance with the search order, the colors and shades of the frame of each window and the background in the window are changed and displayed.

FIG. 76 shows an example of the browse section display window. The black window on the left and right and below the window frame is the browse window displayed in the immediately preceding search process, and the other windows are for the browse displayed in the search process.

(Fifteenth Embodiment) The fifteenth embodiment is directed to the first embodiment.
This is an embodiment in which the processing content of the search result display unit 207 is changed in the embodiment.

The processing other than the display of the search result is the same as that of the first embodiment, and the description is omitted.

FIG. 77 and FIG. 78 show the flow of processing of the search result display section of this embodiment.

The processing started from other components includes generation of a search result display window (step 770).
3), display of search results (step 7704), setting of document propriety information (steps 7705 and 7706), and the like. Since these processes are the same as those in the first embodiment, description thereof will be omitted. In addition, as processing for mouse input, deletion of the search result display window (step 7708), generation of the search result display window (step 7710), re-search based on document availability information (step 7712), change of viewpoint (step 7712) 7714 and 7
715), and the activation of the browse unit (step 7717) is the same as that in the first embodiment, and thus the description is omitted.

Note that in the flowchart shown in the figure,
After the last process of each process, the process waits until an input event such as a mouse input is performed. When an input is performed, the process is repeated from the beginning (step 7703 in FIG. 77).

If the "rotate" command is input, after step 7718, the flow waits for an input event in step 7719. If the mouse input was made near the axis, step 7721 is executed. In step 7721, the point corresponding to each document in the search result is rotated by a predetermined angle θ about the axis designated by mouse input, and the coordinates of each point are newly obtained. Next, step 77 in FIG.
At 22, the coordinates are projected onto a two-dimensional plane that is set in advance. In step 7723, the document is displayed on the display based on the coordinates of each point thus obtained on the two-dimensional plane.

As for the display of the search result, step 77
24 and step 7725 are performed. These steps are Step 7722 and Step 772, respectively.
Same as 3.

In the display of the search result, each point represented by a coordinate system of three or more dimensions is displayed on a display capable of displaying only a two-dimensional coordinate system. Therefore, it is necessary to obtain a coordinate vector of each point on the projected two-dimensional coordinates by projecting each point of the multidimensional coordinate system onto the two-dimensional coordinate system.

A general method of irradiating a point in a coordinate system of a certain dimension to a lower coordinate system is described in, for example, a document (Ichiro Satake: “Linear Algebra”, Shokabo, Mathematics Selection Book 1). A mathematical method has been disclosed as a projection method. 2
Projection onto a dimension is a special case of this, and the method and mathematical formula disclosed in the above-mentioned document may be used.

Also, in the case where the rotation is performed by an angle θ about a certain axis, the coordinate vector a of each point is converted as shown in Expression 21 to obtain a new coordinate vector a ′.
(Equation 21 represents rotation about the z-axis).

[0261]

(Equation 13) After performing the conversion according to Equation 21 for all the coordinates, by performing projection to two dimensions again, the rotated display can be performed.

A general rotation can be expressed as a linear sum of rotations about each axis. Therefore, for general rotation, it is only necessary to sequentially perform conversion regarding rotation about each axis.

FIG. 79 shows the state of the rotation operation.
At the point in time when the display of (a) is displayed, a "rotation" command is input, and further, by clicking the mouse near the vertical axis, the display rotated about the vertical axis as shown in (b) is displayed. It can be carried out.

In step 7725, when each point on the multidimensional coordinates is displayed on a two-dimensional display,
By displaying the auxiliary line (broken line) as shown in FIG. 80, it is possible to transform to express a three-dimensional effect. This auxiliary line is drawn from each point in the yz plane, the zx plane, and the xy plane.
An auxiliary line shown in FIG. 80 can be obtained by obtaining a line segment in each of the x, y, and z axes up to the point where the plane intersects.

In this embodiment, a means for instructing with a mouse is used as an instructing means for performing a rotation operation. In recent years, however, a data glove realized as a data input means for displaying a three-dimensional object is used. It is possible to deform as follows.

The present invention is not limited to the above-described embodiment, but can be implemented with appropriate modifications without changing the gist.

[0267]

According to the present invention, when the user designates with a keyword or a natural language sentence, the input can be analyzed to retrieve a document from a plurality of different viewpoints. To display the degree of similarity with the search command. As a result, the user can intuitively judge the validity of the search result from the display result, and a series of operations for understanding individual documents can be efficiently advanced. Document retrieval can be realized. In other words, the validity of the document can be quickly determined from the search result of the document, so that a series of search operations for understanding each document can be smoothly performed.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of a first embodiment of the present invention.

FIG. 2 is a diagram showing a functional configuration of the first embodiment.

FIG. 3 is a diagram showing a functional configuration of an input analysis unit.

FIG. 4 is a diagram showing a processing flow of an input analysis control unit.

FIG. 5 is a diagram showing a flow of processing of a search instruction generation unit.

FIG. 6 is a diagram illustrating an example of a viewpoint extraction rule.

FIG. 7 is a diagram showing a format of a search instruction and an example thereof.

FIG. 8 is a diagram showing a functional configuration of a search unit.

FIG. 9 is a diagram showing a processing flow of a search control unit.

FIG. 10 is a diagram showing a configuration example of a keyword index.

FIG. 11 is a view showing a format of a search result and an example thereof.

FIG. 12 is a diagram showing a processing flow of a search control unit.

FIG. 13 is a diagram showing a functional configuration of a history management unit.

FIG. 14 is a diagram showing a processing flow of a history management control unit.

FIG. 15 is a diagram showing a format of a search node in a search history storage unit and an example thereof.

FIG. 16 is a diagram showing a flow of processing of a search history display unit.

FIG. 17 is a diagram showing an example of a search history display.

FIG. 18 is a diagram showing a flow of processing of a search result display unit.

FIG. 19 is a diagram showing an example of a search result display.

FIG. 20 is a diagram showing a format of document availability information and an example thereof.

FIG. 21 is a diagram showing a flow of processing of a browse unit.

FIG. 22 is a view showing a storage example of the contents of a document.

FIG. 23 is a diagram illustrating a display example of a browse unit.

FIG. 24 is a diagram showing a processing flow of a search result display unit according to the second embodiment of the present invention.

FIG. 25 is a diagram showing a processing flow of a search result display unit of the second embodiment.

FIG. 26 is a diagram showing an example of a search result display.

FIG. 27 is a diagram showing an example of a search history display according to the third embodiment of the present invention.

FIG. 28 shows an example of a search statement input window;

FIG. 29 is a diagram showing a functional configuration of a fourth embodiment of the present invention.

FIG. 30 is a diagram showing a functional configuration of an input analysis unit.

FIG. 31 is a diagram showing a flow of processing of an input analysis control unit.

FIG. 32 is a diagram showing a flow of processing of a search instruction generation unit.

FIG. 33 is a diagram illustrating an example of a viewpoint extraction rule.

FIG. 34 is a diagram showing a format of a search command and an example thereof.

FIG. 35 is a diagram showing an example of a viewpoint extraction result display.

FIG. 36 is a diagram showing a functional configuration of a search unit.

FIG. 37 is a diagram showing a processing flow of a search control unit.

FIG. 38 is a diagram showing a configuration example of a keyword index.

FIG. 39 is a view showing a format of a search result and an example thereof.

FIG. 40 is a diagram showing a processing flow of a search control unit.

FIG. 41 is a diagram showing a functional configuration of a history management unit.

FIG. 42 is a diagram showing a flow of processing of a history management control unit.

FIG. 43 is a diagram showing a format of a search node in a search history storage unit and an example thereof.

FIG. 44 is a view showing a processing flow of a search history display unit.

FIG. 45 is a diagram showing an example of a search history display.

FIG. 46 is a view showing a display example of search node information.

FIG. 47 is a view showing a processing flow of a search result display section.

FIG. 48 is a view showing the flow of document clustering processing.

FIG. 49 is a view showing a flow of clustering processing.

FIG. 50 is a view showing an example of cluster / document correspondence data.

FIG. 51 is a diagram showing a display example of a document cluster.

FIG. 52 is a diagram showing an example of a search result display.

FIG. 53 is a view showing a format of a document propriety information and an example thereof.

FIG. 54 is a diagram showing an example of a search result display.

FIG. 55 is a view showing an example of a browse section display window.

FIG. 56 is a view showing the flow of processing of a browse unit.

FIG. 57 is a view showing an example of document data.

FIG. 58 is a view showing an example of display of a browse unit.

FIG. 59 is a view showing a display example of a document cluster according to the fifth embodiment of the present invention.

FIG. 60 is a view showing an example of a search result display.

FIG. 61 is a diagram showing a search result display window according to the sixth embodiment of the present invention.

FIG. 62 is a view showing a search result display window of the sixth embodiment.

FIG. 63 is a view showing a window displaying the number of documents including a viewpoint according to the seventh embodiment of the present invention.

FIG. 64 is a diagram showing an example of a numerical expression extraction rule according to the eighth embodiment of the present invention.

FIG. 65 is a diagram showing a search result display window.

FIG. 66 is a diagram showing a search result display window according to the ninth embodiment of the present invention.

FIG. 67 is a diagram showing an example of a viewpoint extraction rule according to the tenth embodiment of the present invention.

FIG. 68 is a diagram showing an example of registration of a viewpoint extraction rule according to the tenth embodiment of the present invention.

FIG. 69 is a view showing a window for a search history display unit according to the eleventh embodiment of the present invention.

FIG. 70 is a view showing an example of a search history display.

FIG. 71 is a view showing an example of a browse section display window according to a twelfth embodiment of the present invention.

FIG. 72 is a view showing an example of a browse section display window.

FIG. 73 is a view showing an example of data in personal information storage means.

FIG. 74 is a view showing a display example of a search result display section window according to the thirteenth embodiment of the present invention;

FIG. 75 is a view showing an example of display of a search result display window.

FIG. 76 is a view showing an example of a browse section display window according to a fourteenth embodiment of the present invention.

FIG. 77 is a view showing a processing flow of a search result display section of the fourteenth embodiment of the present invention.

FIG. 78 is a view showing a processing flow of a search result display section of the fourteenth embodiment of the present invention.

FIG. 79 is a view showing a search result display example 1 in the fourteenth embodiment of the present invention.

FIG. 80 is a diagram showing a search result display example 2 according to the fourteenth embodiment of the present invention.

[Explanation of symbols]

101 central processing means, 102 storage means, 103 display controller, 104 display means, 105 input controller, 106 input means, 201 input section, 202
... input analysis unit, 203 ... search unit, 204 ... history management unit,
205: search history storage unit, 206: document data storage unit
207: search result display section, 208: search history display section, 2
09: browse unit, 301: input analysis control unit, 302:
Search command generation unit, 303: viewpoint extraction unit, 304: morphological analysis unit, 305: syntax analysis unit, 801: search control unit, 8
02 ... index scanning unit, 803 ... adaptive unit, 804 ...
Distance calculation unit, 1301 history management control unit, 1302 search node generation unit.

────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Yumi Mizutani 1 Toshiba R & D Center, Komukai, Kawasaki City, Kanagawa Prefecture (56) References JP-A-3-15981 (JP, A) Shinkai Masafumi, Hitoshi Tabuchi, Yoichi Muraoka, Proposal of a paper retrieval method based on the semantic structure of a subject, Proc. Of the 46th IPSJ Annual Conference (4), Japan, Japan Information Processing Corporation Academic Society, March 1, 1993, pp. 207-208 Takashi Kitagawa, Yasushi Kiyoki, The Mathematical Model of Meaning and its Application to Databases, IEICE Technical Report (DE92-6), Japan, The Institute of Electronics, Information and Communication Engineers, May 22, 1992, VOL. 92, No. 51, p. 43-50 Yukio Nakamoto, Yasuo Tanozaki, Isamu Iwai, Full Text Search Experiment Using Japanese Language Analysis, Proc. Of the 46th IPSJ Annual Convention (3), Japan, Information Processing Society of Japan, March 23, 1993, p. 3-125 − 3-126 Ken-ichi Nogami, Yukio Nakamoto, Masato Yajima, Yasuo Tano, Development of Browsing Function Using Semantic Structure of Documents, IPSJ 45th (1992) National Convention Lecture Paper Shu (3), Japan, Information Processing Society of Japan, October 11, 1992, p. 3-193-3-194 Atsushi Tamura, Concept Mapping System SPRINGS Based on Symbolic Mechanics, Transactions of the Information Processing Society of Japan, Japan, Information Processing Society of Japan, April 15, 1992, Vol. 33, No. 4, p. 465-470 (58) Field surveyed (Int. Cl. ⁷ , DB name) G06F 17/30 JICST file (JOIS)

Claims (10)

(57) [Claims] 1. A document information retrieval apparatus for retrieving a desired document from a document stored in a database, a document data storage unit for storing the document, and input means for inputting a keyword or a natural language sentence for document retrieval. And analyzing the keyword or the natural language sentence from the input means, dividing the sentence into different viewpoints, and extracting from the individual sentences of the document.
Created by merging the independent words issued for each viewpoint
Input analysis means for converting the search command into a search command to be searched, and a search means for calculating a distance between a search target document stored in the document data storage unit and the search command for each viewpoint and obtaining a document having a short distance. A document information search device characterized by the following: 2. A search result display means for displaying a similarity relationship between each document and a search command in a multidimensional manner by assigning a distance between a document and a search command sentence obtained for each view to each view as an axis. 2. The document information search device according to claim 1, comprising: 3. An abstract sentence of a document is displayed by selecting one of the documents displayed in a multi-dimensional manner on a search result display means, and information on important words included in the document is listed. And a means for highlighting the result of the determination of the necessity of the document to be browsed or the document input by the user with respect to the document position in the multidimensional display. 2. The document information search device according to 2. 4. A means for inputting a range of a multi-dimensionally displayed document group as a search result to a search result display means, and enlarging / displaying the multi-dimensionally displayed search result according to the input of the range of the document group. 3. The document information retrieval apparatus according to claim 2, further comprising means for calculating a distance for reduction. 5. Search history storage means for storing a search instruction and a search result as a pair as a search history, and means for displaying the search history in a tree structure in which a pair of the search instruction and the search result corresponds to one node. 2. The document information search device according to claim 1, further comprising means for selecting a node in the tree structure to reproduce a display of a search result corresponding to the node. 6. A search result display means including means for collectively displaying a set of documents at a predetermined distance from a predetermined position among a set of documents displayed as a search result in a multidimensional manner as one cluster. 3. The document information search device according to claim 2, wherein: 7. The document information retrieval apparatus according to claim 6, further comprising means for changing the display of the cluster according to the number of documents included in the cluster in which the document group is collected. 8. The document information retrieval apparatus according to claim 6, further comprising means for designating a cluster in which the document group is put together and collectively giving a result of determining whether or not documents in the corresponding cluster are necessary. apparatus. 9. A window is generated by selecting a document from a group of documents displayed multidimensionally on a search result display means, an abstract of the document is displayed in the window, and information on important words included in the document is displayed. 7. The document information according to claim 6, further comprising: browsing means for listing the documents, and means for arranging the windows in a fixed order according to attributes such as the position and date of the document displayed in the window. Search device. 10. A point corresponding to each document of the search result is rotated by a predetermined angle about a designated axis,
2. The document information search device according to claim 1, further comprising means for obtaining coordinates of each point.